OEE explained: Availability × Performance × Quality
Overall Equipment Effectiveness splits everything a machine loses into three factors that multiply: the time it was not running, the speed it gave up while running, and the output it made but had to throw away. The value of the decomposition is not the single percentage at the end. It is that each letter has a different fix, and picking the wrong one is free to do and expensive to keep doing.
OEE = Availability × Performance × Quality
The three letters, and what each one costs you
They multiply, so a line at 90% on each letter is running at 73% overall — and no single letter looks alarming on its way there.
- Availability — breakdowns and changeovers. The machine is stopped, and you can see it is stopped. Fixed with maintenance policy.
- Performance — minor stoppages and slow running. The machine is on and nothing shows as down. This is the loss operators most often cannot name, because there is nothing to point at.
- Quality — scrap and rework. Output was produced, capacity was spent, and the unit does not ship. Downstream of a constraint it compounds multiplicatively.
Why preventive maintenance is so often the wrong answer
Preventive maintenance raises Availability. If the loss is in Performance, it raises the letter that was never the problem — and charges you upkeep for the privilege. This is the sharpest single statement the decomposition can make, and it is directly testable: in one Flowline puzzle a perfectly balanced line bleeds around 8% of its output to minor stoppages, the canvas never flags a station as down, Availability and Quality both read 100%, and the line still falls short. Preventive maintenance is a proven failing fix there. The stoppage policy is the working one.
Quality behaves differently again. Scrap compounds down a line — measured as exactly the product of per-station yields — so fixing precision only at the bottleneck measurably underperforms fixing it everywhere downstream of the constraint.
What the simulation gets checked against
Flowline’s numbers are checked against closed-form results from the operations-management literature by a test suite that runs on every change, with every expected value recomputed independently from the primitives the simulation itself consumes — never by asking the game’s own readout whether it agrees with itself. For OEE that means availability derived from failure and repair rates lands within ±0.02 of the closed form, the effective process time breakdowns inflate is 0.09% off, the variability term is 0.7% off, and the Performance ratio between stoppage policies is 0.29% off. Yield compounding is exact, because it is a deterministic accumulator.
Try it on a real line
Each of these is a fixed-seed puzzle factory in Flowline’s Challenge Mode. Every one is verified before it ships: the intended fix must win, doing nothing must fail, and a plausible wrong fix must also fail — so the puzzle isolates the idea it claims to teach. They run in your browser.
- The Invisible Thief — Ship apparel at ≥ 49.5/min Availability 100%. Quality 100%. The line still falls short, and the intuitive fix is proven to fail.
- Death by a Thousand Cuts — Ship furniture at ≥ 52/min Yield compounding downstream of the constraint, with a three-way proof that the partial fixes are not enough.
- Three Ways to Lose an Hour — Ship apparel at ≥ 24/min One station losing all three letters at once. Every single-lever and two-lever fix measurably falls short.